As sewbots threaten Asia's sweatshops, we need to decide who will benefit

If only it was ever as easy as data in = results out. The quality, accuracy, granularity, accessibility, and existence of data is an important prerequisite. (For healthcare in America, HIPAA will be a big hurdle) Deep learning has been a big advance in how machine learning has been done, but we have to be careful not to oversell it either or assume it is a panacea to all problems in any domain. You can't just say "I'll just use a deep net here" and expect to be done. Getting meaningful results takes training, proper design of networks, filtering, and feedback systems. Lastly there's the issue of what is unspecified. What if an architecture AI designs an optimal building conforming to all proper checks, but is in the end ugly as sin. You spent all that time creating the perfect algorithm that was trained on architecture rules and experience, but either forgot to include or couldn't derive a metric for aesthetics or uniqueness. What if AI doctors become infallible with instant access to contraindications and medical literature, but in the end has no bedside manner and no physical presence to actually convince a patient to listen to them? What if your tax bot was deemed illegal for helping circumvent paying taxes? You might have a solution for all of these, but the fact that problems like these exist/could exit should be a sign that nothing is "a simple a+b=c".